# SHIP > SHIP is an agentic engineering platform: your autonomous software team. Its team of AI coding agents plans, builds, reviews and tests each change and delivers a pull request with proof it works. SHIP reports what each mission delivered, how long it took, and what it cost. SHIP, at letsship.ai, is a software engineering platform. It is unrelated to the shipping and logistics companies that share the word, such as SHIP.com and Easyship. Assign backlog tickets to SHIP's AI software factory from Linear, Jira, GitHub Issues or a terminal. Its team of AI coding agents opens a pull request in GitHub with proof it works. Teams choose the agents, models, provider contracts, engineering practices, and delivery pipelines used for each role. SHIP is an agent orchestration platform that provides an agentic SDLC for ticket-to-PR delivery. Its autonomous SDLC coordinates specialist agents for planning, building, reviewing, deploying, and testing while engineers retain control over acceptance and exceptions. **The problem** Agents already write a growing share of your production code. It lands faster than teams can verify its quality or account for its cost. Teams pick their own agents, tools, and models, and nothing connects that choice to what ships or the uneven quality it ships at. As agent adoption grows, that gap becomes unreviewed code, unpredictable spend, and outcomes no one can attribute. This is where teams lose control of quality and cost. **The solution** - **Autonomous delivery:** Operate AI agents that review and test as fast as they build, so work ships while your engineers stay in control. - **Cost optimization:** Monitor every step, compare performance, and optimize AI costs confidently. - **AI sovereignty:** Run missions on the models, tools, and AI contracts your organization already uses. SHIP is the only platform that runs coding agents like your team and proves what they ship and cost. **The principles behind SHIP** - **Engineering quality is non-negotiable.** Generated code goes through the same review, test, CI, and acceptance gates as any other change. Faster output is useful only when the result matches the quality expectations. - **Measure outcomes, not consumption.** Token volume is a session stat and a cost input. Cycle time, lead time, delivery frequency, quality, and resolved work tell an engineering team whether the system is improving. - **Humans and agents improve the system together.** Agents take on repetitive delivery work. Engineers set direction, handle exceptions, and improve the practices that make the next run better. - **Customers keep control.** Teams should be able to switch agents, models, and providers without rebuilding their workflow. The platform should help reduce AI costs, not profit from making them larger. Use the full context file for a consolidated, LLM-optimized account of SHIP's product, operating model, security posture, frequently asked questions, founder, and principles. ## Product - [SHIP homepage](https://letsship.ai/): Product positioning, autonomous delivery workflow, AI agent team, proof of work, cost optimization, AI sovereignty, and integration model. - [Full SHIP context](https://letsship.ai/llms-full.txt): Expanded product, company, and FAQ content in one Markdown document. ## Platform Each page describes one way to use SHIP, grouped by who it is for. Pages marked "early access" describe a capability that is opening to early users and is not generally available yet; the rest describe what SHIP does today. - [Platform overview](https://letsship.ai/platform): All the ways to use SHIP, grouped into available now and early access. - [Spec-driven development](https://letsship.ai/platform/spec-driven-development): Available now, for developers. Send a written spec with ship delegate plan and get back a reviewed, tested pull request. - [Agent to PR](https://letsship.ai/platform/agent-to-pr): Available now, for developers. ship delegate pr sends a pull request your coding agent wrote to SHIP. If CI fails, SHIP fixes it. Then SHIP reviews and tests the pull request and gives it back for you to merge. No ticket is needed. - [Code migrations](https://letsship.ai/platform/code-migrations): Available now, for developers. You split a migration into steps, and each step runs as its own mission and comes back as a verified pull request. File each step as an issue or send it with ship delegate plan. - [Incident response](https://letsship.ai/platform/incident-response): Early access, for developers. SHIP reads an alert and answers it with a tested fix. Today, assign the issue an alert opens to SHIP in Linear, or comment @SHIP on it in GitHub Issues. - [Codebase wiki](https://letsship.ai/platform/codebase-wiki): Early access, for developers. A Markdown wiki in your repository, updated after every merge. Today a scheduled task can update your docs. - [Agentic SDLC](https://letsship.ai/platform/agentic-sdlc): Available now, for engineering teams. Assign an issue and get back a verified pull request, built on your own coding agents and models. - [Auto maintenance](https://letsship.ai/platform/auto-maintenance): Available now, for engineering teams. Scheduled missions for dependency upgrades, flaky tests and security advisories. - [Agentic testing](https://letsship.ai/platform/agentic-testing): Available now, for engineering teams. In every mission the Tester checks the acceptance criteria on a preview and attaches a browser trace and screenshots. ship delegate pr --at qa sends it any pull request. - [Multi-agent PR review](https://letsship.ai/platform/pr-review): Early access, for engineering teams. Several reviewer agents on every pull request. Today ship delegate pr reviews and tests one open pull request. - [Security review](https://letsship.ai/platform/security-review): Early access, for engineering teams. A security review on every pull request, with each finding proven on a preview and fixed in a pull request. Today ship delegate pr --prompt-file runs a security review on one pull request. - [Jira](https://letsship.ai/platform/jira): Early access, for engineering teams. Assign a Jira ticket to SHIP and get back a reviewed, tested pull request. - [Slack](https://letsship.ai/platform/slack): Early access, for engineering teams. Start a SHIP mission from a Slack thread and get the status back in the thread. Today SHIP works on Linear and GitHub Issues, and ship delegate plan builds from a spec. - [Production vibe coding](https://letsship.ai/platform/vibe-coding-for-production): Available now, for founders. Your AI assistant or coding agent sends work to SHIP through the skill, the CLI or the MCP server. - [Governed prototyping](https://letsship.ai/platform/governed-prototyping): Early access, for founders. Product and design teams build prototypes that engineering reviews and merges. - [App templates](https://letsship.ai/platform/app-templates): Early access, for founders. Start a new product from a template in your own repository. Scheduled missions then upgrade dependencies, fix flaky tests and patch security advisories. - [Content factory](https://letsship.ai/platform/content-factory): Early access, for founders. Each week SHIP turns search and AI-answer data into a list of pages you approve, and writes, reviews and tests each one as a pull request. Today you can send a written brief with ship delegate plan. Every platform page is also published as plain markdown for agents at https://letsship.ai/platform//md, for example https://letsship.ai/platform/agentic-sdlc/md. ## Documentation - [Documentation](https://letsship.ai/docs): What SHIP is, the two ways work is handed over, and how a mission runs from plan through build, review, preview deployment, and test to a pull request. - [Getting started](https://letsship.ai/docs/getting-started): Setup by pointing a coding agent at the hosted skill, the token and account it needs, and a first mission to run end to end. - [Ways of working](https://letsship.ai/docs/getting-started/ways-of-working): The two paths into the same pipeline. Ticket to pull request for people who work in the tracker, agent to pull request for developers handing work over from a terminal, and how one project runs both. - [Integrations](https://letsship.ai/docs/getting-started/integrations): Connecting an outside assistant or service. Which surface to use, what a token needs to carry, and why a follow-up instruction is not a restart. - [Human in the loop](https://letsship.ai/docs/configuration/human-in-the-loop): Checkpoints that hold a run at a stage you want to read, answering a hold, and the controls for steering or stopping a mission. - [Deployments](https://letsship.ai/docs/configuration/deployments): How a mission gets a live preview environment to test against, deployed by the project's own pipeline with its own credentials. - [Retry limits](https://letsship.ai/docs/configuration/retry-limits): How long the agents keep working a problem before handing it back, and what an oscillation is. - [Agent prompts](https://letsship.ai/docs/configuration/agent-prompts): Extending how each role behaves with instructions from your own repository, layered on top of SHIP's charter for that role. - [Use your repository's MCP servers](https://letsship.ai/docs/configuration/mcp-servers): Let agents use the servers in your .mcp.json. Bind each one once per project, and its credential never enters the sandbox. - [Connect a hosted MCP server](https://letsship.ai/docs/configuration/mcp-servers/hosted-servers): Bind GitHub's server with a token header and Linear's with an OAuth sign-in. A server that starts with Docker gets a hosted endpoint. - [Run an MCP server in the sandbox](https://letsship.ai/docs/configuration/mcp-servers/servers-in-the-sandbox): Start servers that need no secret with npx or uvx, as the repository declares them. They serve the planner and builder. - [Isolate an npm MCP server that needs a secret](https://letsship.ai/docs/configuration/mcp-servers/isolated-servers): Run the npm package outside the sandbox with a stand-in for its key. The real key goes only to the hosts you allow. - [Fix an MCP server that doesn't work](https://letsship.ai/docs/configuration/mcp-servers/troubleshooting): Read what each server status means and what a stopped mission says. See which servers an agent had. - [ship.yml schema reference](https://letsship.ai/docs/configuration/schema): Every field of the per-project manifest, generated from the same definition the platform validates against. - [ship.yml JSON Schema](https://letsship.ai/schema/ship.yml.json): The machine-readable manifest schema, for editor validation and CI. - [CLI](https://letsship.ai/docs/cli): Handing work over from a terminal with ship delegate, instruct, and status, including per-mission harness, model, and prompt overrides. - [API](https://letsship.ai/docs/api): The HTTP API behind the console, generated from the API's own definition, with authentication and the scope sets worth granting. - [MCP server](https://letsship.ai/docs/mcp): The hosted MCP server at mcp.letsship.ai. Client configuration, the tools it exposes, and a worked hand-off from an assistant that never sees your repository. - [A2A agent](https://letsship.ai/docs/a2a): SHIP as an A2A 1.0 agent. Another agent hands it a task and follows the task to a verified pull request. The Agent Card is at https://letsship.ai/.well-known/agent-card.json. - [Agent setup skill](https://letsship.ai/SKILL.md): The one-line setup for a coding agent. It installs the CLI and a SHIP skill for each CLI capability. - [Agent skills index](https://letsship.ai/.well-known/agent-skills/index.json): Every SHIP skill, with its URL and its SHA-256 digest. - [Agent authentication](https://letsship.ai/auth.md): How an agent gets a SHIP token from its user and sends it to the API and the MCP server. - [Documentation for agents](https://letsship.ai/docs/llms.txt): Every documentation page in one file, each with its one-line description. ## Research - [SWE-in-a-team benchmark](https://letsship.ai/blog/swe-in-a-team): 13 coding agent configurations (harness and model) each given the same 20 tickets and taken through plan, build, CI, review, preview deploy and test, graded by hidden acceptance tests, with the per-run data published as CSV. - [Codex vs Claude Code](https://letsship.ai/blog/codex-vs-claude-code): Codex resolved 55 of 60 missions at $3.63 per resolved ticket, and claude-code 58 of 60 at $4.69. ## Guides and definitions - [Agentic SDLC, defined](https://letsship.ai/blog/agentic-sdlc): What an agentic SDLC is, how it differs from traditional, AI-driven and autonomous SDLC, what happens when a gate rejects a change, and how to measure cycle time, cost per change and resolve rate. - [What is an AI software factory](https://letsship.ai/blog/what-is-an-ai-software-factory): The parts an AI software factory needs, how the term differs from its older DevOps meaning, and how to assemble one around Claude Code or run it on SHIP. - [How to build an AI software factory](https://letsship.ai/blog/how-to-build-an-ai-software-factory): A roadmap in four stages: assess, enable, orchestrate and operate. Each stage lists when it starts, what it needs and the milestone that ends it. - [What is agentic coding](https://letsship.ai/blog/what-is-agentic-coding): What agentic coding and agentic engineering are, and the gates and retry limits that make agent output safe to merge. - [What is a dark software factory](https://letsship.ai/blog/what-is-a-dark-software-factory): Where the term comes from, what 260 agent missions show machine checks miss, and how dark SHIP can run. - [Claude Code on Linear issues](https://letsship.ai/blog/claude-code-on-linear-issues): Assign a Linear issue to SHIP and get a planned, CI-checked, reviewed and preview-tested pull request from the claude-code harness. - [Codex on Linear issues](https://letsship.ai/blog/codex-on-linear-issues): Assign a Linear issue to SHIP and get a reviewed, preview-tested pull request from the codex harness. - [Claude Code on Jira](https://letsship.ai/blog/claude-code-on-jira-issues): Connect Claude Code to Jira through Atlassian's remote MCP server. SHIP runs its loop on Jira tickets: plan, CI, review and a preview test. - [Claude Code in Slack](https://letsship.ai/blog/claude-code-in-slack): Set up Claude Code in Slack or Claude Tag today. SHIP's loop for a Slack thread (plan, CI, review, preview test) is in early access, and today it runs on Linear and GitHub Issues. - [What is governed vibe coding](https://letsship.ai/blog/what-is-governed-vibe-coding): What governed vibe coding is, what a team must control before a vibe-coded change ships, and what SHIP runs today. - [AI prototype to production](https://letsship.ai/blog/ai-prototype-to-production): Take an AI prototype to a reviewed, tested pull request from an issue, a spec or its own pull request. A workspace for non-engineers is in early access. - [Vibe coding security](https://letsship.ai/blog/vibe-coding-security): The risks security guides name for vibe coding, the check that catches each one, and which checks SHIP runs today. - [What is shadow AI](https://letsship.ai/blog/shadow-ai-in-software-development): Shadow AI defined, what it looks like in software development, and a sanctioned path from an AI-built prototype to a merge. - [Codex as reviewer for Claude Code](https://letsship.ai/blog/codex-reviewer-for-claude-code): Keep claude-code on the builder and put codex on the reviewer, in ship.yml or for a single mission, and compare reviewers from per-stage records. - [Run the pi coding agent headless](https://letsship.ai/blog/run-pi-coding-agent-headless): Run pi as SHIP's builder in an isolated sandbox on an open-weight model through your own OpenRouter key, with review and QA around it. - [Run Claude Code headless](https://letsship.ai/blog/run-claude-code-headless): Run Claude Code in an isolated sandbox that never holds your Anthropic key, with CI, review and QA around it. - [Run Codex headless](https://letsship.ai/blog/run-codex-headless): Run codex as SHIP's builder or reviewer in a sandbox with no OpenAI key inside, and CI, review and QA around it. - [MCP servers for coding agents](https://letsship.ai/blog/mcp-servers-for-coding-agents): Give every pipeline agent the MCP servers your repository declares. Servers run remote, isolated or in the sandbox. Credentials stay outside the sandbox, and review and QA get read-only tools. - [GitHub and Linear MCP servers for coding agents](https://letsship.ai/blog/github-linear-mcp-servers-coding-agents): Bind GitHub's server with a scoped token and Linear's with an OAuth sign-in. Keep review and QA read-only, and see each call on the run. ## Compare - [Factory AI alternatives](https://letsship.ai/compare/factory-ai-alternatives): Factory, Devin, Warp Factories, Tembo, OpenHands and SHIP compared on where work starts, whose model keys pay for inference, which coding agents run, what checks a change and how cost is reported, each claim sourced to the product's own pages. - [Claude Code alternatives](https://letsship.ai/compare/claude-code-alternatives): Eight terminal coding agents compared with Claude Code on license, models, headless mode and price. - [Devin alternatives](https://letsship.ai/compare/devin-alternatives): Eight cloud agents compared with Devin on where work starts, models, checks and price. - [Background coding agents](https://letsship.ai/compare/background-coding-agents): Ten cloud agents that return a pull request, compared on machines, triggers and price. Every article above is also available as plain Markdown by appending `/md` to its URL. ## Use cases - [Use case catalog](https://letsship.ai/use-cases): 224 backlog slices an agent can take end to end, grouped into 29 categories across design, architecture, development, data, test, security, tech debt, operations, delivery, and documentation. Each one carries a goal, acceptance criteria, the evidence that lands on the pull request, and a prompt that opens the ticket. - [Use case catalog for agents](https://letsship.ai/use-cases/llms.txt): Every use case in one file, grouped by lifecycle phase and category, each linking to a plain-markdown rendition at /use-cases//md. ## Company - [Introducing SHIP](https://letsship.ai/blog/introducing-ship): Why SHIP exists: AI adoption got ahead of agentic engineering practice, why teams should measure the work delivered rather than the tokens spent, and how a strong delivery loop changes which model a team needs. - [About SHIP](https://letsship.ai/about): What SHIP does, how it compares with Lovable, Devin, Factory and CodeRabbit, who uses it, the team behind founder Önder Ceylan, how it works, key facts, and FAQ. - [SHIP FAQ](https://letsship.ai/faq): Answers about supported agents and models, subscription keys, secrets, source code, cost, approvals, failure handling, cloud and CI, and coding assistants. ## Contact - [Book a demo](https://letsship.ai/contact/book-a-demo): Schedule a conversation about using SHIP in your engineering organization.